Search Results for "scoring"

X For You Feed Algorithm is the open-sourced core recommendation system that powers the For You feed on X (the social network formerly known as Twitter), and it represents one of the first times a major social platform has published production-level ranking code for public review and experimentation. The repository contains the full pipeline that ingests user engagement and content candidate data, processes it through retrieval, hydration, filtering, scoring, and selection layers, and ultimately ranks posts to show what appears in a user’s feed. At its heart, the system uses a transformer-based model adapted from xAI’s Grok architecture to predict probabilities for various user actions (such as likes, replies, reposts, clicks, and negative signals), then combines those into a weighted final score that drives ranking.

...The algorithm uses features such as sentence length, keyword frequency, and position within the document to determine which sentences are most relevant. By combining these features with a simple scoring mechanism, it produces summaries that are both readable and informative. Originally inspired by research and earlier implementations, textteaser provides a lightweight solution for summarization without requiring heavy machine learning models. It is particularly useful for developers, researchers, or content platforms seeking a simple, rule-based approach to article summarization.